JPH06165387A - System stabilizer - Google Patents

Abstract

PURPOSE:To realize optimum control without as operators determining the threshold for determining the quantity of power limitation, in a system stabilizer. CONSTITUTION:This stabilizer is equipped with quantity-of-electricity detecting means 1 and 5 which detect the output and current of generators G1 to G3 operating in parallel and the voltage of the high-voltage bus of a substation. If a fault occurs in a system, a generator condition operating means 6 computes the inner phase angle of the angular velocity of a generator in fixed cycles. And, an energy operating means 7 computes the kinetic energy of the generator and the position energy in fixed cycles. Moreover, a power limitation judging means 8, where a neural network is applied, judges whether to limit the power of the generator or not. Based on the result, a control means 9 issues a parallel off command for the applicable generator so as to parallel off it, whereby the step-out of a substation is prevented.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system in which a phenomenon in which a plurality of generators are out of step due to a system accident is stabilized by disconnecting a part of the generators from the system. It relates to a stabilizing device.

[0002]

2. Description of the Related Art Conventionally, as a method of calculating the power supply limit amount, that is, the number of disconnected generators, which has been applied to this type of system stabilizer, for example, a generator output before a system accident and a cutoff The main method was to find the number of disconnected generators based on a preset limit calculation curve (threshold value) based on accident information obtained from equipment information and the like. FIG. 9 is a diagram illustrating a conventional system stabilizing device. The conventional device has a power detection unit 1 that detects the total output Po of the generator before the accident, and an accident type determination unit 2 that determines the accident type based on the breaker information (opening information of CB6 and CB7) after the occurrence of the system accident. When,
Generator total output Po before the accident detected by the power detection unit 1
And the accident type determined by the accident type determination unit 2, the control amount calculation unit 3 that calculates the control amount from the control amount calculation curve as shown in FIG. 10 and obtains the optimum number of off-line generators.
And a control unit 4 for shutting off the generator.

[0003]

In the above-mentioned conventional apparatus, some threshold value (for example, the control amount calculation curve) is necessary for determining the power source limit amount, which is usually decided by the operator of the apparatus through trial and error. It was being done. In addition, since the control accuracy of the entire device largely depends on this threshold value, the operator of the device, who is the setting person of the device, takes a great responsibility. Further, since the threshold value is usually determined based on the offline digital simulation result for the most severe system condition, there is a drawback that it is over-controlled in almost all other system conditions. The present invention has been made in view of the above circumstances, and it is not necessary for an operator to set a threshold value for determining a power supply limit amount by trial and error, and optimal control is possible over all system states. It is intended to provide a system stabilizing device.

[0004]

In order to achieve the above object, in a system stabilizing device according to the present invention, the output and current of a generator operating in parallel at a power plant and the voltage of a high voltage busbar of the power plant are detected. Inputting the electric quantity detected by the electric quantity detecting means, and calculating the state quantity of the generator after the accident in a constant cycle on condition that an accident has occurred in the system Machine state calculating means, energy calculating means for inputting the calculated generator state quantity and the detected output of the generator, and calculating the kinetic energy of the generator and the potential energy of the system in a fixed cycle; Input a predetermined number of time-series data at the closest point in the calculated kinetic energy and potential energy and the detected output of the generator in the power station before the accident. There was constructed from the power supply limit determining means for determining whether to power limiting the generator, a control means for disconnection from the system to the appropriate power generator on the basis of the determination result.

In the system stabilizing device described above, when an accident occurs in the system, the generator state calculating means determines the angular velocity of the generator based on various electric quantities detected by the electric quantity detecting means at a constant cycle of about 10 msec. The internal phase angle is calculated. Then, when it is detected that the accident has been eliminated due to changes in the generator output, etc., and after a certain period of time elapses, power is generated using the angular velocity and internal phase angle of the generator calculated in the fixed cycle, and the detected output of the generator. The kinetic energy of the machine and the potential energy of the system are calculated at regular intervals. Then, based on the time-series data of a predetermined number of these points at the closest time point and the output before the accident of each generator detected by the electricity amount detection means, power generation is performed by the power supply limitation determination means applying a neural network. Determine whether the power source of the generator in the station should be restricted,
Based on the result, the control means outputs a disconnection command to the corresponding generator to disconnect the generator, thereby preventing a step out of the power plant.

[0005]

Embodiments will be described below with reference to the drawings. Figure 1
Shows an example of the configuration of the system stabilizing device according to the present invention. In FIG. 1, the same parts as those in FIG. 9 are designated by the same reference numerals and the description thereof is omitted. In FIG. 1, 5 is a voltage / current detection unit for detecting the total current of the power plant and the voltage of the high-voltage bus BUS,
Inputting the electric power P, the voltage V, and the current I from each of the detection units 1 and 2, 6 is used to calculate the angular velocity and the internal phase angle of the target generator at a fixed cycle of 10 to 20 msec when an accident occurs in the system. The generator state calculator 7 uses the internal phase angle of the angular velocity of the generator calculated by the generator state calculator 6 and the generator output detected by the power detector 1 to determine the kinetic energy of the generator and the system An energy calculation unit that calculates the potential energy in a fixed cycle of 10 to 20 msec, 8 is a power supply limitation determination unit that applies the concept of a neural network, and is the closest point in time between the kinetic energy and the potential energy calculated by the energy calculation unit 7. A predetermined number of time-series data and the outputs of G1 to G3 before the accident detected by the power detection unit 1 are input to determine whether or not the power sources of the generators G1 to G3 should be restricted. A control unit 9 outputs a disconnection command (cutoff command) for disconnecting the generator from the grid based on the disconnection information of the generator calculated by the power supply limitation determination unit 8.

Here, an outline of the neural network will be described. The neural network (NN) is
It can be said that the neural network of the biological system was realized on a computer. So far, NNs with various structures have been proposed,
Layered nets are excellent in pattern recognition, and non-layered nets are suitable for optimization problems. The layered NN suitable for this problem changes its structure through learning using learning data, and outputs a reasonable result even for input data other than learning data. From this point as well, it can be said that the NN is suitable for the problem of utilizing knowledge that is difficult to stabilize and rule.

FIG. 2 shows a three-layer NN. Output layer from top,
It is called the middle layer and the input layer. The number of nodes (neurons) in each layer is determined according to the target problem. The mathematical model of each neuron is shown in FIG. One neuron has k inputs. On the other hand, the output V _j
Is as follows. Where W _ji is the weight for each input, θ _j is the threshold,
I _j is an external input, and f _i is a non-linear function called a neuron characteristic.

As learning data, consider a set of an input value and a resulting output value. In the initial state, appropriate values are set for the weights and thresholds of the NN neurons.
When this input data is input to the NN, some value is output from the output layer. Naturally, this value is different from the output value of the learning data. Therefore, the weight and the threshold value are changed so that the two match. As this method, Ru
Backpropagation by melhart can be mentioned.
In this way, by learning the NN using a large number of learning data, it is possible to output an appropriate value even for an input different from the learning data.

Next, a method of calculating the angular velocity ω of the generator and the internal phase angle δ input to the NN will be described. First, a method of calculating the initial phase angle with the main system will be described. In the target system of Fig. 1, the driving point impedance when looking at the main system from the high-voltage bus (BUS) of the target power plant is jX _S, and the generator current when all generators in the power plant are equivalent to one If I _G , the target system of FIG. 1 is modeled as shown in FIG. In FIG. 4, V _t is the bus voltage of BUS,
jX _t and jX _q are the internal impedance of the generator and the impedance of the step-up transformer when the generators G1 to G3 in the power plant are regarded as one equivalent (hereinafter referred to as equivalent G).
Based on the phase of the bus voltage V _t of BUS, the internal voltage V _G of the equivalent _G and the voltage V _S of the main system side are respectively Required from. Therefore, the initial internal phase angle δ _GS of the equivalent G when the voltage V _{S of the} main system is used as the phase reference is as shown in the vector diagram of FIG. Becomes

Next, a method of calculating the angular velocity ω and the phase angle δ after the accident has occurred will be described. Let M _{T be} the sum of the inertia constants of the generators G1 to G3, and let P _T (o) be the sum of the generator outputs before the accident. Next, if the generator output value obtained by sampling at Δt intervals after the accident occurs is P _T (t), the equivalent G angular velocity ω
(t) and internal phase angle δ (t) can be obtained as the solution of the following motion method.

Next, the kinetic energy of the generator and the potential energy of the system, which are the input data of the NN, will be described. An energy function method is one of the methods for determining the stability of a power system. In this method, the stability is determined by comparing the kinetic energy stored in the generator due to a three-phase short circuit accident and the potential energy of the system after the accident is eliminated. Equation (9) below shows an example of the energy function. In equation (9), the first term on the right side is the kinetic energy of the generator, and the second term on the right side is the potential energy of the system. Where δ _c is the internal phase angle of the generator at the moment of accident elimination, P
_m is the mechanical input to the generator and P _e is the generator output.

FIGS. 6 (a) and 6 (b) show the kinetic energy (hereinafter, referred to as kinetic energy of the generator when a three-phase short circuit accident occurs in the power system.
KE) and potential energy of the system (hereinafter referred to as PE) are shown as an example of temporal movement. Fig. 6 (a) is stable, Fig. 6 (b) is unstable (generator It is a case where a part is out of sync.). Now, as a feature of the energy method,
It is known that the total energy obtained by adding KE and PE is constant, and as can be seen from Figs. 6 (a) and 6 (b), the energy accumulated by the grid fault is KE after the fault is cleared.
And PE share each other. When stable, the system absorbs the KE in the accident, so PE increases and KE decreases. On the other hand, in the case of instability, the system does not have the ability to absorb KE in an accident, so KE increases and PE decreases, resulting in instability. Therefore, the energy (KE) of a short time (for example, 100 msec) after the accident is removed can be applied to a neural network that has learned such features of the energy method in advance.
And PE), it is possible to quickly and appropriately determine the stability and calculate the power supply limitation amount.

Specific energy value calculation formulas are formulas (10) and (11). Here, M _T is the total value of the inertia constants of the generators G1 to G3,
_{P T (o), P T} (t k) are each accident before, which is the total output of the generator G1~G3 after the accident occurred. Also, ω
(T _k), [delta] (t _k) are (7), angular velocity and the internal phase (8) equivalent G calculated from the equation (generator G1~G3 equivalent generator when contracted to an aircraft) It is a horn. Here, the operation of the NN used in the power supply limitation determination unit 8 will be described. FIG. 7 shows the configuration of the NN. The number of neurons in the input layer is (the number of time series data of the kinetic energy KE of the generator calculated by the energy calculation unit 7 and the potential energy PE of the system)
+ (The number of generators operating in the power station), for example, if the number of time-series data is 5 and the number of generators is 3, 5
X2 + 3 = 13. Also, the number of neurons in the output layer is necessary for the number of generators in order to determine whether or not there is a power source restriction for each generator. The effects on the NN are as follows.

The NN is the most effective index for judging the stability, time series data of the kinetic energy of the generator and the potential energy of the system, and the pre-accident of each generator necessary for judging the power source limited generator. When the output is input, the neurons in the input layer, the intermediate layer, and the output layer calculate their own output from the equation (1) using the input data or the output of the previous layer neuron. The outputs of the input layer and intermediate layer neurons are input to the neurons of the intermediate and output layers, respectively. On the other hand, the output of the output layer neuron is a determination result of whether or not the power source of each generator should be limited, and takes a binary value of 0, 1. Output is "0"
In case of 1, it is determined that power supply is not required, and in case of 1, it is determined that power supply is limited.

An embodiment of the present invention based on the above concept will be described below with reference to FIG. First, in a steady state before a system fault, the power detection unit 1 detects the output of each of the generators G1 to G3 operating in parallel at the target power plant, and the generator state calculation unit 6, the energy calculation unit 7, and the power supply limitation determination. Send to Part 8. On the other hand, in the voltage / current detection unit 5, the generators G1 to G
The total current of 3 and the voltage of the high voltage bus BUS of the power plant are detected and sent to the generator state calculation unit 6. In the generator state calculation unit 6, the initial internal phase angle δ _GS of the power plant equivalent G is (2), (3) as an initial value for calculating the phase angle after the occurrence of the accident, based on the input electric quantity. ), (4), and the outputs of the generators G1 to G3 are summed and stored as the equivalent G output P _T (o) before the accident. Similarly, the energy calculation unit 7 sums the outputs of the generators G1 to G3, and outputs the output G _T of the equivalent G before the accident P _T (o) and the power supply limitation determination unit 8 to the outputs of the generators G1 to G3 to NN. It is stored as the output P ₁ (o), P ₂ (o), and P ₃ (o) before the accident that is input.

On the other hand, when the occurrence of an accident in the system is detected by the change in the electric power of the target power plant, the electric power detection unit 1 outputs Δt.
The outputs of the generators G1 to G3 are detected and sent to the generator state calculator 6 and the energy calculator 7 at regular intervals. In the generator state detection unit 6, the detected outputs of the generators G1 to G3 are summed to P _T (t _k ), and the angular velocity ω after the accident occurs by using the equations (6), (7) and (8). (T _k ), phase angle δ (t _k ).
Is calculated and the result is sent to the energy calculator 7. The energy calculation unit 7, the sum of the output of the generator G1~G3 that the detected search of P _T (t _k), generator status computing section 6
Angular velocity ω (t _k ) and phase angle δ (t _k ) sent from
And P that was calculated and stored in the previous step
The kinetic energy KE (t _k ) of the generator and the potential energy PE (t _k ) of the system are calculated from Eqs. (10) and (11) using _T (t _k-1 ) and δ (t _k-1 ). , And sends the result to the power supply limitation determination unit 8. The power supply restriction determination unit 8 stores this result as time series data of kinetic energy KE and potential energy PE.

Next, when the fact that the accident has been eliminated is detected by a change in the output of the power plant or the like and a certain period of time has passed, the power supply limitation determination unit 8 indicates the nearest time point in the stored time series data of PE and KE. Predetermined number and generator G detected before the accident
Based on the pre-accident outputs of 1 to G3, the presence or absence of the power source limitation of the generators G1 to G3 is determined by the action of the NN described above. If the neuron in the output layer corresponding to each generator outputs “1”, it is determined that power supply limitation is necessary and this result is sent to the control unit 9. The control unit 9 disconnects the corresponding generator based on this result, thereby stabilizing the remaining generators in the power plant.

Finally, the learning method of the NN will be described.
As a learning data collecting means, a transient stability analysis widely used for determining the stability of a power system is used off-line. The transient stability analysis is for simulating the dynamic behavior of the generator when an accident occurs in the system, in which the output of the generator for calculating the input data to the NN used in the present invention, the target Includes total power plant current and high voltage bus voltage. Therefore, by using these data, the kinetic energy of the generator and the potential energy of the system, which are the input data to the NN, can be easily obtained.
The pre-accident output of each generator can be used as it is as input data to the NN. Further, since the stability can be easily determined by looking at the simulation result, the teaching signal can be easily obtained. Therefore, the learning data can be easily collected according to the flowchart shown in FIG.

That is, in step S1, an accident point or an accident mode assumed to be a system condition to be operated is set. Next, in step S2, a transient stability analysis without power supply limitation is performed, and in step S3, the output of each target generator, the total current of the target power plant, and the voltage of the high voltage bus bar are stored. Next, in step S4, using the transient stability analysis results stored in step S3, the equations (2) to (8) and (10) and (11) are used.
From the equation, the kinetic energy KE _{i of the} generator (i = 1, ...,
m) and potential energy PE _{i of the} system (i = 1, ..., m)
To calculate. This and the pre-accident output P _j (o) (j = 1, ..., m) of each generator stored in step S3 are input data to the NN. Further, in step S5, the transient stability analysis when the power source is limited is performed until the target power plant becomes stable, and the generator for which the power source must be limited for stabilization is obtained. Then, based on this result, a teaching signal is created and stored in step S6. For example, of the generators G1 to G3, G2
If the result is that only the power supply needs to be limited, the teaching signal for this case is (0, 1, 0). When power supply restrictions for G1 and G3 are required (1, 0, 1)
Becomes

A plurality of learning data sets can be collected by changing the system condition, assumed accident, and accident mode and repeating the process of FIG. By using this collected learning data set to learn the weights and thresholds by backpropagation, the operator does not have to make trial and error to modify the NN for the system stabilizing device with sufficient accuracy. , Can be built. In the system stabilizing device according to the present embodiment, it is not necessary to determine the threshold value for determining the power supply limitation amount by trial and error, and the responsibility of the device operator is reduced. Further, by including all possible system conditions, assumed accidents, and accident modes in the learning data of the NN, it is possible to obtain an extremely highly accurate system stabilizing device.

[0021]

As described above, according to the present invention, it is not necessary for the operator to determine the threshold value for calculating the power supply restriction amount, and it is possible to provide a system stabilizing device with extremely high accuracy. it can.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of a system stabilizing device according to the present invention.

FIG. 2 is a diagram showing a general configuration of a neural network.

FIG. 3 is a diagram illustrating a mathematical model of each neuron.

FIG. 4 is a configuration diagram when the power system shown in FIG. 1 is modeled.

5 is a vector diagram showing the relationship between V _G and V _B in FIG.

FIG. 6 is a diagram showing a temporal movement of energy when a three-phase short circuit accident occurs in the power system.

FIG. 7 is a block diagram of a neural network used in the present invention.

FIG. 8 is a flowchart showing a method for learning a neural network.

FIG. 9 is a configuration diagram showing a conventional system stabilizing device.

FIG. 10 is a control amount calculation curve diagram applied to a conventional system stabilizing device.

[Explanation of symbols]

 1 Power Detection Section 5 Voltage / Current Detection Section 6 Generator Status Calculation Section 7 Energy Calculation Section 8 Power Limit Judgment Section 9 Control Section

Claims (1)

[Claims] 1. A system stabilizing device applied to a power system in which a power station having a plurality of generators installed in parallel and a main system are connected via a power transmission line, the parallel operation is performed at the power stations. The electricity amount detecting means for detecting the output and current of the generator and the voltage of the high voltage busbar of the power plant, and the electricity amount detected by the electricity amount detecting means are input, and an accident has occurred in the system. Kinetic energy of the generator by inputting the generator state calculation means for calculating the state quantity of the generator after the accident in a constant cycle under the condition When a predetermined number of kinetic energy and potential energy calculated by the energy calculation means are used at the nearest point by using an energy calculation means for calculating the potential energy of the system at a constant cycle and a neural network. Inputting the sequence data and the detected output of the generator in the power station before the accident, a power supply limitation determining unit that determines whether or not the power source of the generator should be limited, and a corresponding generator based on the determination result. And a control means for disconnecting the system depending on the system.